An Improved Framework for the Analysis and Dissemination of Seismic Site Characterization Data at Varying Resolutions
- Author(s): Ahdi, Sean Kamran
- Advisor(s): Stewart, Jonathan P
- et al.
The most commonly used parameter for representing site conditions for ground motion studies is the time-averaged shear-wave velocity in the upper 30 m, or VS30. While it is preferred to compute VS30 from a directly measured shear-wave velocity (VS) profile using in situ geophysical methods, this information is not always available. One major application of VS30 is the development of ergodic site amplification models, for example as part of ground motion model (GMM) development projects, which require VS30 values for all sites.
The first part of this dissertation (Chapters 2-4) addresses the development of proxy-based models for estimation of VS30 for application in subduction zone regions. These procedures are applied at 6433 strong motion recording stations (SMSs) for the NGA-Subduction project, which has the goal of developing GMMs for global subduction zone (SZ) earthquakes. Relatively detailed VS30 prediction models are developed in this thesis for application to the Cascadia SZ in the Pacific Northwest (PNW) region and the Alaska/Aleutian SZs. These are the portions of the United States at greatest risk to seismic hazards from subduction zone earthquakes. In these regions, only 8% of SMSs have measured VS30 values.
Proxy-based VS30 statistical models based on in situ measurements were developed from information including surficial geology, topographic gradient (i.e. “slope”), and geomorphic terrain categories. The PNW and Alaska studies result in proxy-based VS30 models based on (1) a hybrid of generalized surficial geologic groups conditioned on topographic slope, and (2) geomorphic terrain categories. With 928 measured VS30 values available in the PNW, statistically robust proxy models were developed, with 18 generally well-populated geology groups assigned logarithmic mean and standard deviation VS30 values, six of which are conditioned on topographic slope, and certain groups reflecting glacial and volcanic geology. Additionally, 13 of a possible 16 terrain classes were well-populated, and these were also assigned VS30 statistical moments. Ultimately, due to strong correlation between the two proxies but an overall lower dispersion of model residuals for the hybrid geology-slope proxy compared to the terrain proxy, use of the hybrid slope-geology proxy model was recommended. For Alaska, a different approach was required, as most geology groups were not well populated. In these cases, VS30 data from the PNW were borrowed for similar geologic groups, residuals analysis for Alaska-only and combined Alaska-PNW group moments were computed, and bias was considered in model prediction. The standard deviations of such groups’ model predictions were increased by adding an epistemic uncertainty in the mean to reflect the uncertainty in adopting a proxy for use outside of its original intended region of application.
Other SZ regions included in NGA-Subduction project rely on regional VS30 prediction models developed previously or concurrent with this project by others (i.e., Japan, Taiwan, New Zealand, and Chile), or required development of procedures as part of the present work to assign VS30 where regional models are unavailable (i.e., Central America/Mexico and South America outside of Chile). The VS data collection effort for the latter two regions resulted in a general lack of measured VS30 data and proxy information, precluding robust proxy model development. As such, an existing geomorphic terrain class proxy model was borrowed from California, with an additional assigned epistemic uncertainty in the mean to account for increased uncertainty in implementing proxy-based models outside of their original intended region of use. Basin depth terms are also provided for VS profiles that exceed velocity thresholds (e.g., 1.0 or 2.5 km/s) or estimated for regions where 3D seismic velocity models exist (e.g., Cascadia, Japan, New Zealand, and Taiwan).
A similar study was undertaken in Iran to populate a site database with VS30 values for a ground motion modeling. Analysis of measured Iranian VS data and comparison of within-group moments for geology and terrain proxies in other regions around the world showed that the Iranian VS30 values did not vary much across different geology groups, a possible sign that the seismic refraction velocity data lacked adequate resolution in the upper 30 m to provide accurate VS30. To mitigate this, a third approach for proxy development was formally developed, in which moments for similar geologic groups were borrowed from the PNW and California, averaged, and used for assignment to Iranian strong motion stations, again with care to factor in inter-regional epistemic uncertainty. This work defined the framework for the assignments of VS30 to the aforementioned data-poor regions of Central America, Mexico, and South America.
The second part of this dissertation (Chapter 6) concerns the development of the United States Community VS Profile Database (PDB), a major multi-institutional effort to develop an open-access VS profile database for sites in the United States. The data described herein was collected from diverse sources that include consulting engineering reports from private industry, university research reports and other documents, federal open-file and similar reports, California state agency documents, and reports provided by electric utilities for selected sites. All data are strictly within the public domain, but much of it was for practical purposes inaccessible to most potential users. The VS data sources encompass a wide array of geophysical techniques, are presented in many different formats, and are accompanied by widely divergent supplementary data, including P-wave velocities, geotechnical logs and other data, and penetration test data. A relational database schema of sufficient breadth and flexibility was developed to accommodate this diverse data set. The data are digitized and otherwise prepared in the standardized format specified by the database schema. A web interface (www.uclageo.com/VPDB) was developed for data query, visualization, and download. This resource is anticipated to be useful to geotechnical engineers and engineering seismologists for diverse applications in research and industry practice.